Abstract
The perceived direction of a translating tilted line shows strong biases for motion along non-cardinal axes (up to 40°, Loffler & Orbach, 2001). Are these biases restricted to single lines? Do they vanish if additional motion information is present?
Methods: Two high-contrast, intersecting (intersection visible or occluded) lines (4.85x0.25dva) translated for 195ms with average speed of 5deg/s. 16 different absolute directions were tested (0 to 360). Different cross shapes were created by varying the angle between the two lines (20°, 45°, 90° and 120°). The axis of the cross was aligned with the direction of motion (symmetrical) on some trials and misaligned in others.
Results: For crosses, as for single lines, there are no biases for motion along the cardinal axes. Biases are reduced but not eliminated for oblique motions (up to 20°). Biases with two lines are different to that observed with a single dot (which usually yields the smallest bias), even when the intersection of the two lines is visible and could be tracked. The bias for crosses depends on the angle between the lines and the angle between their axis and the direction of motion. Generally, they are smaller when the lines are symmetrical around the direction of motion. In this case, a narrow angle (20°) shows the smallest bias (≤10°). Perception changes when the two lines are not arranged symmetrically and is biased towards the line closest in orientation to the direction of motion of the pattern.
Conclusions: Against expectation, strong biases persist even where a host of motion signals are available. In addition to the veridical signals from each of the line-endings and the intersection of the lines, the combination of two 1-D signals from each line segment permits, in theory (IOC), veridical motion estimation. Our results show that neither strategy is used successfully for oblique directions of motion.
Supported by EPSRC grant GR/S59239/01 to GL